[R] Plotting (mixed) line types and legends

[Prof Brian Ripley]

On 28 Jul 2003, Peter Dunn wrote:

> I have a question about plotting line types and
> legends.  Here's a short piece of code to
> demonstrate:
> 
> x <- y <- seq(1,10)
> plot(x,y, type ="l", lty="33")
> lines(x,y+1,  lty=1)
> legend( 8,2,legend=c("lty=1","lty=\"33\""), lty=c(1,"33") )
> 
> On my system (see below), the line types in the legend
> are not the same as in the plot--in particular, lty=1.
> (In the plot, the  lty=1  is solid as expected; in the
> legend, it is dots close together.)
> 
> Is this because I shouldn't mix different ways of defining
> line types in the legend?  (ie stick to lty=c(1,2) only, 
> for example)?  If so, I couldn't find it documented.

Yes.  You have actually use lty=c("1", "33") by the coercion rules: the
argument you passed has to be one mode and that mode is character.
lty=c("solid", "33") works.

Whether lty="1" should work is a moot point: it seems device-dependent.
It appears to work on postscript() and windows(), so looks like a bug in 
the X11() device (which I presume you used, although you did not say so).
On the other hand, one could argue it is a bug that it works anywhere
as it is not as specified in ?par.

> Any explanation appreciated.  If anyone know hows to
> do such a thing (ie define different line types in the
> legend), I'd love to hear.  (And yes, I need to do it--the
> example above was a short snippet that just demos the problem.)

You have not demonstrated that, and I believe you could always use the 
character-string form of lty.


-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595